Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 FULL PAPER Open Access Contributions of poroelastic rebound and a weak volcanic arc to the postseismic deformation of the 2011 Tohoku earthquake Yan Hu1*, Roland Bürgmann1, Jeffrey T Freymueller2, Paramesh Banerjee3 and Kelin Wang4 Abstract A better understanding of fluid-related processes such as poroelastic rebound of the upper crust and weakening of the lower crust beneath the volcanic arc helps better understand and correctly interpret the heterogeneity of postseismic deformation following great subduction zone earthquakes The postseismic deformation following the 2011 Mw9.0 Tohoku earthquake, recorded with unprecedented high resolution in space and time, provides a unique opportunity to study these ‘second-order’ subduction zone processes We use a three-dimensional viscoelastic finite element model to study the effects of fluid-related processes on the postseismic deformation A poroelastic rebound (PE) model alone with fluid flow in response to coseismic pressure changes down to and 16 km in the continental and oceanic crusts, respectively, predicts to cm uplift on land, up to approximately 20 cm uplift above the peak rupture area, and up to approximately 15 cm subsidence elsewhere offshore PE produces up to approximately 30 cm of horizontal motions in the rupture area but less than cm horizontal displacements on land Effects of a weak zone beneath the arc depend on its plan-view width and vertical viscosity profile Our preferred model of the weak sub-arc zone indicates that in the first years after the 2011 earthquake, the weak zone contributes to the surface deformation on land on the order of up to 20 cm in both horizontal and vertical directions The weak-zone model helps eliminate the remaining systematic misfit of the viscoelastic model of upper mantle relaxation and afterslip of the megathrust Keywords: Poroelastic rebound; Weakened lower crust beneath the arc; Giant earthquake; Subduction zone; Viscoelastic postseismic deformation; Finite element model; Numerical simulation Background Geodetic observations of deformation before, during, and after M ~ megathrust earthquakes illuminate the mechanics and rheology of the subduction zone system Wang et al (2012) summarized three primary subduction processes that dominate earthquake cycle deformation following a great megathrust earthquake: aseismic afterslip on the subduction thrust, viscoelastic relaxation of the upper mantle, and re-locking of the fault Immediately after the earthquake, afterslip on the subduction megathrust and a transient viscoelastic response of the mantle result in rapidly decaying trench-ward surface displacements (e.g., Pollitz et al., 2008; Ozawa et al., 2012; * Correspondence: yhu@seismo.berkeley.edu Berkeley Seismological Laboratory and Department of Earth and Planetary Science, University of California Berkeley, 307 McCone Hall, Berkeley, CA 94720, USA Full list of author information is available at the end of the article Lin et al 2013) Decades after the earthquake, the coastal area moves towards the land due to the re-locking of the fault, while viscoelastic relaxation of the mantle still causes prolonged seaward motions in the inland area (e.g., Hu et al., 2004; Wang et al., 2003, 2007; Suito and Freymueller, 2009; Hu and Wang, 2012) Later in the earthquake cycle (e.g., McCaffrey et al., 2013), the earthquake-induced stresses in the mantle are mostly relaxed, and the effects of the re-locking of the fault dominate leading to a landward displacement gradient consistent with elastic deformation about the subduction thrust coupled in the upper approximately 50 km of the lithosphere (Savage, 1983) The recent devastating M ~ megathrust earthquakes in Sumatra, Chile, and Japan provide unique opportunities to improve our understanding of the subduction earthquake cycle through observations of the deformation with modern space-geodetic techniques © 2014 Hu et al.; licensee Springer This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 Here we focus on modeling the postseismic deformation following the 11 March 2011 Mw9.0 Tohoku earthquake in NE Japan (Pollitz et al., 2011; Ozawa et al., 2012; Iinuma et al., 2012), exploring the role of fluids in earthquake cycle deformation Specifically, we consider (1) the contribution of fluid flow in response to coseismic pressure changes in the lithosphere to the postseismic deformation and (2) the role of fluids rising from the subducting slab in the volcanic arc of NE Japan in producing localized weakening of the lower crust Tens of meters of instantaneous coseismic slip of the fault cause sudden pressure changes in the surrounding rocks Pore fluid pressure immediately increases in the compressional areas and decreases in dilatational areas in the initial undrained condition After the earthquake, fluids will migrate from high-pressure areas to lowpressure areas resulting in time-dependent surface deformation associated with poroelastic rebound (Peltzer et al., 1996, 1998) Migration of fluids thus causes the pore fluid pressure to evolve towards an equilibrium condition in which the earthquake-induced fluid flow has completed, commonly referred to as ‘drained’ condition This time-dependent process (e.g., Jónsson et al., 2003; Masterlark, 2003) is controlled by the variable viscosities of fluids, the rock properties, and the complex permeability structure of the lithosphere A common way to predict the deformation resulting from the completed poroelastic rebound is to consider only the difference in elastic coseismic deformation between the undrained condition immediately after the earthquake and the fully relaxed equilibrium condition long after the earthquake (e.g., Masterlark, 2003; Jónsson et al., 2003) This is accomplished by differencing coseismic deformation models in which portions of the lithosphere where earthquake-induced fluid flow is believed to occur are modeled with undrained and equilibrium values of Poisson's ratio Study of poroelastic rebound helps to better understand the contributions of fluidflow processes in shaping the transient stress field and evolving earthquake hazard (e.g., Peltzer et al., 1998; Hughes et al., 2010) and to gain insights on the permeability/porosity structure of and fluid flow in subduction zone systems (e.g., Nur and Walder 1990) The poroelastic rebound model has been applied to study crustal deformation associated with subduction zone earthquakes such as the 2004 Mw9.2 Sumatra (Hughes et al., 2010) and 1980 Mw8.0 Jalisco-Colima, Mexico, earthquakes (Masterlark, 2003) Hughes et al (2010) presented a finite element model of the poroelastic rebound following the 2004 Sumatra earthquake that produced up to a few tens of centimeters of horizontal displacements near the trench and less than 30 cm uplift in the vicinity of the rupture zone Masterlark (2003) suggests that a model with bulk permeability of the Page of 10 oceanic crust less than 10−17 m2 may explain the quasistatic coupling of an earthquake swarm that has a 63-day lag time following the 1980 Mw8.0 Jalisco-Colima earthquake However, the contribution of poroelastic rebound to the postseismic deformation of the 2011 Tohoku earthquake has yet to be investigated (Ozawa et al., 2012; Johnson et al., 2012; Diao et al., 2014) It is also known that compaction and heating of the hydrated subduction slab results in fluids migrating into the overlying mantle wedge (Manning, 2004) These fluids weaken the overriding plate and may cause partial melting (e.g., Saffer and Bekins, 1999; van Keken et al., 2002) Through modeling heat flow, seismic tomography, and magnetotelluric data, Muto (2011) and Muto et al (2013) have proposed that viscosities of the lower crust below the arc in NE Japan are several orders of magnitude lower than in the surrounding crust After examining interseismic strain anomalies and the coseismic deformation of the 2011 earthquake in NE Japan, Ohzono et al (2012b) proposed a weak zone below the tens of kilometers wide Ou-backbone range, in the vicinity of the arc A low-viscosity lower crust (2 to × 1018 Pa s) at depths >20 km is also indicated by the postseismic relaxation of the 2008 Iwate-Miyagi Nairiku earthquake located in the arc (Ohzono et al., 2012a) Postseismic deformation following the 2011 Mw9.0 Tohoku earthquake has been recorded at more than 1,200 continuous land Global Positioning System (GPS) stations (Ozawa et al., 2012) as well as a few marineacoustic campaign GPS stations (Sato et al., 2013; Kido et al., 2013; Japan Coast Guard and Tohoku University 2013; Watanabe et al 2014) at unprecedented high spatial and temporal resolutions The 2011 earthquake thus provides a unique opportunity to study processes other than the three primary deformation processes mentioned above, illuminating the role of fluids and material heterogeneity in the postseismic deformation We believe that it is important to understand the possible contributions of these higher-order effects to the postseismic deformation field as they will impact any postseismic deformation models which parameterize structure and properties of the Earth through comparing with observations In this paper, we present a three-dimensional (3D) viscoelastic finite element model to illuminate the effects of the poroelastic rebound in the crust and the rheology heterogeneity below the arc Methods Geodetic observations and postseismic displacement estimates Postseismic displacements at geodetic stations are estimated based on the land GPS observations and seafloor GPS-acoustic (GPS-A) measurements (Figure 1) We obtained daily time series of more than 1,200 continuous Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 Page of 10 Figure Tectonic setting and postseismic GPS observations in NE Japan (a) Horizontal displacements Red arrows represent 2-year GPS observations since the 2011 Tohoku earthquake The solid magenta circle represents the location of the example GPS station whose time series is presented in Additional file 1: Figure S1 Solid black triangles represent active volcanoes (b) Vertical displacements of GPS stations years after the 2011 earthquake Black contours at 5-meter intervals show the coseismic slip distribution from Iinuma et al (2012) GPS stations (GEONET) processed in ITRF2008 (Altamimi et al., 2011) by the Geospatial Information Authority of Japan (GSI) (Miyazaki et al 1998) The GPS time series span from as early as 1996 to March 2013 The GPS time series represent a combined signal of non-tectonic seasonal deformation, interseismic locking, and postseismic processes In this work, we are interested in deformation only due to postseismic processes Estimates of the postseismic deformation directly from the daily GPS time series suffer from the epoch noise level We take the following steps to estimate the total postseismic displacements over a 2-year period (from 12 March 2011 to 30 March 2013) This approach is thus not comprised by any data gaps or problems at the time exactly years after the earthquake First, we select an interseismic time window in which previous earthquakes have minimum contributions to surface deformation A function consisting of a linear trend and seasonal sinusoidal terms is fitted to the interseismic time series to approximate the pre-earthquake trends to account for non-tectonic seasonal deformation and the interseismic locking (Additional file 1: Figure S1) We subtract the pre-earthquake motions from the postseismic time series to estimate postseismic displacements only due to the earthquake-related processes that are examined in this work We fit a parametric model to the time series and evaluated the model to provide displacements over desired time windows Finally, displacements at all stations are referenced to station FUKUE (station ID 950462) such that displacements at these stations are comparable to model-predicted results that are with respect to the fixed upper plate For details of processing of the GPS time series, please see Additional file 1: Section Land GPS stations recorded up to approximately m horizontal and approximately 1.2 m vertical postseismic displacement within years after the 2011 earthquake (Figure 1) All the GPS stations move in roughly the same seaward direction as during the coseismic rupture (Figure 1a) Two years after the earthquake, the eastern coastal stations landward of the rupture zone feature up to approximately 20 cm uplift while areas farther inland and north experienced up to approximately 15 cm subsidence (Figure 1b) In addition to the GEONET data, we also consider 2-year postseismic displacements at six GPS-A stations that were repeatedly surveyed by the Japanese Coast Guard, starting to weeks after the earthquake (Japan Coast Guard 2012; Japan Coast Guard and Tohoku University 2013; Watanabe et al 2014) (Figure 1) The GPS-A station displacements are also relative to station FUKUE The elastic strain associated with subduction of the Pacific plate at a rate of approximately cm/year (e.g., Sella et al., 2002; Apel et al., 2006) makes a modest contribution to the large postseismic displacements at these sites (Sato et al., 2013) Because of the campaign-mode observations of the GPS-A stations, we not take the same steps as in processing the daily time series of the GEONET stations (Additional file 1: Section 1) Effects of interseismic locking are accounted for by removing the interseismic velocities of those marine stations reported by Sato et al (2013) from the postseismic Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 displacements Figure shows displacements of these GPSA stations only due to postseismic processes Figure 1a illustrates that MYGI and KAMS have moved landward while the other stations are moving seaward Except for station CHOS that exhibits insignificant vertical deformation, all the other offshore stations underwent subsidence of approximately 10 to 40 cm in the first years after the 2011 earthquake (Japan Coast Guard and Tohoku University 2013) At FUKU and MYGW, more than 50% of the 2-year subsidence took place in the first months, while stations KAMN, KAMS, and MYGI experienced a more gradual decay of the subsidence rate Finite element model The finite element model used in this work is based on previous mechanical models developed to study the postseismic and interseismic deformations of the Sumatra, Chile, and Cascadia margins (Hu et al., 2004; Wang et al., 2012; Hu and Wang, 2012) The finite element model includes an elastic 40-km-thick upper continental plate, an elastic 80-km-thick subducting slab, and viscoelastic continental and oceanic upper mantles (Figure 2a) Poroelastic rebound in the shallow crust and a weak volcanic arc (gray-shaded areas in Figure 2a) will be investigated in the ‘Poroelastic rebound in the crust’ and ‘Weakened zone beneath volcanic arc’ sections, respectively The bottom of the model is at 500-km depth in the transition zone Lateral model boundaries are set to be at least 1,000 km from the rupture zone Deformation at the model boundaries, except at the free upper surface, is free in the tangential directions and fixed in the normal direction The bi-viscous Burgers rheology, incorporating a transiently relaxing Kelvin solid and steady-state Maxwell fluid, is assumed to represent the constitutive properties of the viscoelastic upper mantle (Bürgmann and Dresen, 2008) Coseismic slip (Iinuma et al., 2012) (Figure 2c) is modeled as sudden forward slip of the megathrust through the split-node method (Melosh and Page of 10 Raefsky, 1981) Note that details of the coseismic source model are not important for the far-field deformation, and different source models yield approximately the same postseismic viscoelastic deformation at the land GPS stations Time-dependent, stress-driven afterslip away from the rupture zone is modeled through a 2-km-thick weak shear zone attached to the megathrust (brown and green layers in Figure 2a) The viscosity of the shallow shear zone (≤50 km, brown layer in Figure 2a) is one order of magnitude lower than that of the deep shear zone (50 to 120 km, green layer in Figure 2a) to produce more afterslips at shallow depths as indicated by observed aftershocks and repeating earthquakes (Uchida and Matsuzawa, 2013) This paper focuses on the effects of fluid-related processes during the early postseismic relaxation First, we present the results of a reference model (REF) with fixed viscoelastic parameters that were based on previous studies (e.g., Hu et al., 2004; Hu and Wang, 2012; Wang et al., 2012) and were found to provide a good firstorder fit to the early postseismic deformation Then we evaluate the impacts of poroelasticity and mantle heterogeneity in the arc center In REF, the shear moduli for the elastic plates and viscoelastic upper mantle are assumed to be 48 and 64 GPa, respectively Poisson’s ratio and rock density are assumed to be 0.25 and 3.3 g/cm3, respectively, for the entire domain The Maxwell steadystate viscosity ηM of the mantle wedge and oceanic mantle is 1019 and 1020 Pa s, respectively ηM of the shallow (≤50 km) and deep (50 to 120 km) afterslip shear zones are 1017 and 1018 Pa s, respectively The Kelvin transient viscosity ηK of the Burgers body in the reference and all the following test models is assumed to be one order of magnitude lower than ηM Details of the reference model and a thorough exploration of the model parameter space will be published elsewhere (Hu et al., manuscript in preparation) Figure Conceptual model parameterization and finite element mesh (a) The finite element model Dark and light gray-shaded regions represent the poroelastic layers and the weak volcanic arc that are considered in the ‘Poroelastic rebound in the crust’ and ‘Weakened zone beneath volcanic arc’ sections, respectively μ, ηM, and ηK are shear modulus, steady-state Maxwell viscosity, and transient Kelvin viscosity, respectively (b) Central part of the finite element mesh Red and black dots represent locations of the land and marine GPS stations, respectively Thick white lines represent coast lines (c) Central part of the finite element mesh with the upper plate removed Color contours are the coseismic slip distribution (Iinuma et al., 2012) Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 Following the approach of developing the FEM mesh in Hu and Wang (2012), we manually derived 32 latitudeparallel profiles based on published slab geometry data (Nakajima and Hasegawa, 2006; Zhao et al., 2009), relocated seismicity (Engdahl et al., 1998), and locations of the trench (Bird, 2003) and the arc Our slab geometry is similar to that used in Iinuma et al (2012) These latitude-parallel profiles were then used to construct the finite element mesh It consists of 147,867 nodal points in 17,408 27-node quadratic elements The element size is on the order of 100 m near the fault and up to 500 km farther away The central part of the mesh is shown in Figure 2b The parallel modeling finite element code PGCvesph was developed at the Pacific Geoscience Centre, Geological Survey of Canada (e.g., Hu and Wang, 2012; Wang et al., 2012) Results and discussion A comparison of the GPS observations with the REF model displacements is presented in Figure REF predicted 2-year displacements fit the first-order pattern of the seaward motion of the land GPS stations (Figure 3a) The systematic misfit of horizontal displacements south of 37° N and along the coast near 40° N may be due to local processes such as aftershocks in this region The subduction of the Philippine Sea plate that is not considered in this work may also contribute to the misfit in the south In the vertical component, REF successfully predicts uplift along the eastern coast behind the rupture zone and subsidence further inland (Figure 3b) REF produces approximately 10 cm subsidence at stations Page of 10 KAMN, KAMS, and MYGI, a pattern consistent with GPS-A observations (Watanabe et al 2014) At MYGW and CHOS, REF underestimates the observed vertical motion At FUKU, the vertical motion predicted by REF is contrary to the observation Horizontal displacements produced by REF are overall consistent with these of GPS-A stations except the directions of KAMS and MYGI Below we explore a series of forward models of (1) the poroelastic rebound of the continental and oceanic crusts and (2) the viscous relaxation of a localized, fluidweakened zone below the NE Japan volcanic arc and explain how deformation from these processes affects the fit of REF to the GPS observations Through these models, we aim to better understand the uncertainties of the model parameters and the role of fluid-mediated processes in the postseismic deformation Poroelastic rebound in the crust In this section, we present test models of poroelastic rebound (PE) in the continental and oceanic crusts Laboratory and geologic studies indicate that crustal permeability decreases rapidly below about 4-km depth (Manning and Ingebritsen, 1999) Based on geothermal models and properties of metamorphic rocks, Manning and Ingebritsen (1999) reported that the permeability in the upper 15 km of the crust decreases logarithmically with a depth from 10−14 to 10−18 m2 Masterlark (2003) proposed that a model with a permeability of the oceanic crust 10−17 m2 well explained the 63-day lag time of an earthquake swarm following the 1995 Mw8.0 Jalisco-Colima mainshock, which is consistent Figure Comparison of GPS observations with reference model-predicted displacements (a) Horizontal displacements Red arrows represent 2-year GPS displacements since the 2011 earthquake Blue arrows are model-predicted displacements Contours are the same coseismic slip distribution (Iinuma et al., 2012) as in Figure (b) Vertical displacements Colored contours are observed 2-year vertical displacements Red and blue arrows represent model-predicted uplift and subsidence Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 with months of observed PE and well water level changes following two Mw6.5 earthquakes in basaltic crust of South Iceland (Jónsson et al., 2003) Therefore, it may take only a few tens of days for the shallow poroelastic layer to relax from the earthquake perturbation Although PE is a complicated time-dependent process, we use a 3D elastic model (the same structure as shown in Figure 2a but the material is elastic) to simulate two end-member states to estimate the total effects of PE The first end-member case represents the immediate response to the earthquake, which is conventionally called the ‘undrained’ condition The second scenario represents the state at which the earthquake perturbation on pore fluid pressure reaches an equilibrium state, and transient poroelastic fluid flow has completed For convenience, we call the second state the ‘equilibrium’ condition to avoid the confusion of the ‘drained’ condition that implies no change in pore fluid pressure because of slow loading processes and high permeability The difference of coseismic model results of these two states thus approximates the total effects of the time-dependent PE that is not modeled in this work Following previous studies of poroelastic rebound (e.g., Masterlark and Hughes, 2008; Hughes et al., 2010), the top layers of the subduction slab and the continental crust are assumed to be poroelastic at the time scales considered here Thicknesses of the poroelastic layer in the slab and continental crust are initially assumed to be 16 and km, respectively Based on previously published studies (summarized in Additional file 1: Table S1), we assume that Poisson's ratio in the continental poroelastic layer is υu = 0.34 under undrained conditions (right after the earthquake) and υ = 0.25 under equilibrium conditions In the oceanic poroelastic layer υu = 0.31 and υ = 0.25 The shear moduli of the continental and oceanic poroelastic layers are 15 and 20 GPa, respectively, Page of 10 for both undrained and equilibrium conditions The magnitude of the difference in Poisson's ratio between these two conditions is likely an upper bound estimate (Additional file 1: Table S1) This poroelastic model thus represents a maximum estimate of PE contributing to the postseismic deformation Tests of depth variation of the shear modulus and Poisson's ratio are detailed in Additional file 1: Section and show that allowing poroelastic fluid to flow deeper in the lithosphere does not substantially change the pattern of the predicted surface deformation The tendency of fluids to flow from high-pressure areas to low-pressure areas causes uplift above and radial displacements away from the rupture zone as illustrated in Figure PE only in the oceanic crust produces surface displacements mostly in a narrow zone close to the trench (Figure 4b) while PE only in the continental upper plate produces displacements across a broader zone (Figure 4a) Note that the sudden decay of coseismic slip from tens of meters to zero near the trench is probably not physical and the resultant large subsidence in this area may be a model-produced artifact (Figure 4) Nevertheless, most significant deformation in either case takes place in the immediate vicinity of the rupture zone Varying the depth extent of the poroelastic layer affects deformation mainly offshore but has little impact for deformation on land (Additional file 1: Figures S4 and S7) PE in the whole continental mantle (e.g., Ogawa and Heki, 2007) has negligible contribution to the surface deformation (Additional file 1: Figure S4d), while PE in the whole oceanic mantle produces up to 20 cm subsidence near the landward edge of the rupture zone and more than 10 cm landward motion near the trench (Additional file 1: Figure S7d) Magnitude and location of the uplift and subsidence produced by PE strongly Figure PE in the upper crust (a) PE only in upper km of continental crust (b) PE only in upper 16 km of the oceanic plate (c) PE in both continental and oceanic crusts, i.e., the combined effect of (a) and (b) Colored contours and black arrows are vertical and horizontal displacements, respectively Hu et al Earth, Planets and Space 2014, 66:106 http://www.earth-planets-space.com/content/66/1/106 depend on source models (Additional file 1: Figure S6) The combined effects of PE in both the upper plate and the slab result in up to approximately 20 cm uplift in the peak rupture area and up to approximately 15 cm of subsidence elsewhere offshore (Figure 4c) Re-equilibration of fluid pressures assuming end-member poroelastic properties produces total horizontal displacements of approximately 30 cm near the offshore rupture area but